Poor Convergence for Transient Subsonic Diffuser sims with flow separation

TWaung · April 15, 2009, 20:06

Hi,

I am modeling a diffusing air inlet that sees high subsonic velocities (0.3 - 0.9). The flow separates off the inlet walls under certain flow conditions that I need to model.

So far, none of my high resolution steady state sims has converged, regardless of how I refine or adjust the mesh. I think this is due to the inherently transient nature of the separated regions.

I have therefore switched to transient simulations. However, I am finding that I need to use a very small timestep size (1.0e-7 s) for the simulation to converge. Using a timestep size of 1.0e-7 the courant number ranges between 0.01 - 0.9. I am using the results from a converged Upwind scheme solution as my initial conditions.

As I increase the time step size, the solution does not converge, and at a certain point, the solver fails (usually preceeded by a huge max Mach number warning). The courant number increases greatly as I increase time step size and the solver fails at a time step size of about 1.0e-4.

I believe part of the reason is that the inlet geometry I am using has a relatively sharp lip. For the flow conditions I am modeling, the local velocity around the lip is very high relative to the freestream (but still subsonic). This high velocity combined with the small cell size around the lip contributes to a very high courant number.

I would like to use a larger time step since I don't have a long time to run these simulations. Using the current time step size it will take months to run a sim.

Does anyone have any suggestions that may allow for the use of a larger time step size?

Some details of my simulation are provided below

Domain Settings
Steady state
Using SST turbulence model, Ideal Gas, Total Energy model

Boundary conditions
- subsonic velocity inlet
- opening pressure outlet. Using the entrain setting since I expect there to be regions of separated flow crossing the outflow plane.

Solver Settings
Time step size: 1.0e-7 s
Advection scheme: High Resolution
Transient Scheme: 2nd order backward Euler

Any suggestions would be greatly appreciated!

Thanks!

ghorrocks · April 16, 2009, 09:25

Hi:

Some ideas:
1) In a steady state run, use local timestep factor=5.0 or so until convergence looks good and finish off with some physical timescale iterations.
2) Extend the inlet and/or outlet lengths if there is recirculations through the inlet/outlet. This will make convergence much more difficult.
3) Have you tried these ideas: http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Glenn Horrocks

April 15, 2009, 20:06	Poor Convergence for Transient Subsonic Diffuser sims with flow separation	#1
TWaung New Member Join Date: Mar 2009 Posts: 14 Rep Power: 17	Hi, I am modeling a diffusing air inlet that sees high subsonic velocities (0.3 - 0.9). The flow separates off the inlet walls under certain flow conditions that I need to model. So far, none of my high resolution steady state sims has converged, regardless of how I refine or adjust the mesh. I think this is due to the inherently transient nature of the separated regions. I have therefore switched to transient simulations. However, I am finding that I need to use a very small timestep size (1.0e-7 s) for the simulation to converge. Using a timestep size of 1.0e-7 the courant number ranges between 0.01 - 0.9. I am using the results from a converged Upwind scheme solution as my initial conditions. As I increase the time step size, the solution does not converge, and at a certain point, the solver fails (usually preceeded by a huge max Mach number warning). The courant number increases greatly as I increase time step size and the solver fails at a time step size of about 1.0e-4. I believe part of the reason is that the inlet geometry I am using has a relatively sharp lip. For the flow conditions I am modeling, the local velocity around the lip is very high relative to the freestream (but still subsonic). This high velocity combined with the small cell size around the lip contributes to a very high courant number. I would like to use a larger time step since I don't have a long time to run these simulations. Using the current time step size it will take months to run a sim. Does anyone have any suggestions that may allow for the use of a larger time step size? Some details of my simulation are provided below Domain Settings Steady state Using SST turbulence model, Ideal Gas, Total Energy model Boundary conditions - subsonic velocity inlet - opening pressure outlet. Using the entrain setting since I expect there to be regions of separated flow crossing the outflow plane. Solver Settings Time step size: 1.0e-7 s Advection scheme: High Resolution Transient Scheme: 2nd order backward Euler Any suggestions would be greatly appreciated! Thanks!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
fluid flow fundas	ram	Main CFD Forum	5	June 17, 2000 22:31
Sphere in a subsonic flow	Rameez	Main CFD Forum	0	September 15, 1999 21:51
Flow visualization vs. Calculated flow patterns	Francisco Saldarriaga	Main CFD Forum	1	August 3, 1999 00:18
Question on 3D potential flow	Adrin Gharakhani	Main CFD Forum	13	June 21, 1999 06:18
computation about flow around a yawed cone	Tylor Xie	Main CFD Forum	0	June 9, 1999 08:33

April 16, 2009, 09:25		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	Hi: Some ideas: 1) In a steady state run, use local timestep factor=5.0 or so until convergence looks good and finish off with some physical timescale iterations. 2) Extend the inlet and/or outlet lengths if there is recirculations through the inlet/outlet. This will make convergence much more difficult. 3) Have you tried these ideas: http://www.cfd-online.com/Wiki/Ansys...gence_criteria Glenn Horrocks